What did the Trump administration argue in its brief supporting OpenAI in The New York Times’ copyright lawsuit, why does the government believe training large language models on copyrighted material generally qualifies as fair use and is important for scientific progress, nation

The administration’s filing is a strong policy endorsement of OpenAI’s fair-use position, but it does not decide the case or bind other courts. It raises the stakes in AI copyright litigation by framing broad access to training data as a U.S. competitiveness and national-security interest. 3

• What the government argued: The Justice Department told the Manhattan court that training LLMs on copyrighted works will generally be fair use because the models use the material to learn statistical relationships and generate new outputs, rather than to republish the works themselves. It cautioned that a model’s outputs that infringe, or training practices involving illicitly obtained material, remain separate issues. 3

• Why it says this matters: The filing says a robust U.S. AI industry is important to scientific advancement, economic opportunity and prosperity, and national security; the administration argues that requiring permission or payment for every work used in training could impede U.S. development relative to foreign competitors. 3 Associate Attorney General Stanley Woodward characterized the policy objective as sustaining a competitive U.S. AI sector. 13

• Legal effect and wider implications: This was a DOJ statement of interest, not a ruling: Judge Sidney Stein must independently apply copyright law and the four fair-use factors. Still, it offers defendants a significant federal-government view in a litigation wave involving OpenAI and Microsoft, as well as Anthropic, Meta and other developers; the eventual appellate rulings, rather than this brief, will establish binding precedent. 13

• The Times’ allegation: The Times alleges OpenAI and Microsoft copied and used millions of its articles without authorization to train models, then created products that can compete with or reproduce Times material—undermining the value of its journalism and licensing market. OpenAI’s public position is that its models are trained on publicly available data and that the practice is protected by fair use. 36

• Why “transformative” remains unsettled: The first merits-stage federal rulings have not produced a clean, universal rule. In the Anthropic book case, Judge William Alsup called training on lawfully acquired books “quintessentially transformative,” while treating acquisition and retention of pirated copies as a distinct potential infringement. 114 Other decisions have placed more weight on proof of market substitution or harm, meaning that a court may regard the analytic use in training as transformative yet still examine whether it damages actual or emerging licensing markets. 112

• The competing responses: The Times’ position is that technological progress should not permit companies to take valuable reporting without consent or compensation. OpenAI says fair use enables models trained on public material and has defended the practice while entering some voluntary content-licensing deals. 6 Woodward emphasized AI competitiveness; Commerce Secretary Howard Lutnick expressed the complementary policy formulation: countries should permit AI training on creators’ work while finding ways to protect artists. 25

The core unresolved question is therefore not simply whether AI is innovative: it is whether wholesale copying for model training, and any resulting output or market substitution, satisfies fair use on the particular evidence in each case. 13